Project

General

Profile

Actions
Copy link

Bug #7790

closed

Kernel panic when creating ZFS pools on CEPH RBD devices

Added by Chris Dunlop about 10 years ago. Updated almost 10 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
-
Target version:
-
% Done:

0%

Source:
Support
Tags:
Backport:
Regression:
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

Creating a ZFS pool on top of krbd causes a kernel panic.

From the ZFSonLinux bug tracker (https://github.com/zfsonlinux/spl/issues/241):

"It seems the rbd driver will overwrite memory in some cases when passed odd-sized bio requests from ZFS. I gave up trying to track down the memory corruption because I couldn't get anything reproducible so I compared ZFS' block I/O to that of other file systems (ext4 and xfs) and discovered that the others only presented nice evenly-sized requests (usually 8 sectors or multiples of 8) but that ZFS presents "unusual" sizes such as 5, 13 and 15 sectors. Apparently the rbd driver doesn't like some aspect of these requests and the result is memory corruption. I've not yet determined whether the transfer size or the buffer alignment is the problem."

Actions
Copy link

Also available in: Atom PDF